The present invention pertains to a method and apparatus for classifying articles according to their color.
In the past, many color sensing principles have been proposed. One such color sensing principle is described in U.S. Pat. No. 4,830,501 to Terashita et al., which consists of measuring light reflected from an article at different wavelengths to detect at least two light reflection values, the difference therebetween or the ratio thereof being indicative of the color of this article. The difference or the ratio of reflection values is then compared to a predetermined threshold value for classification purposes based on color characteristics. These different wavelengths used to detect the reflection values are selected according to the light reflection curves in terms of the reflected light wavelength spectrum, which are specific to the articles about which the color has to be sensed. In order to maximize color resolution, these wavelengths are typically chosen such that, for a given range of colors characterizing these articles, a maximum variation of reflected light intensity values corresponding to the wavelengths is observed. Therefore, wavelengths are generally selected to correspond to some peaks on light reflection curves, and such peaks should be characterized by relatively high amplitude variations depending on the color of the articles, so as to properly carry out color sensing and classification. Typically, each chosen peak wavelength corresponds to a dominant color component which characterizes the color of the articles to be inspected, the relative amount of such a color component being considered as a basis for color sensing and classification.
Other patents such as U.S. Pat. No. 4,204,950, to Burford, U.S. Pat. No. 4,146,135, to Sarkar, reissue No. 29,031, to Irving, and U.S. Pat. No. 3,206,022 to Roberts propose the use of characteristic wavelengths which are selected from light reflection curves for different kinds of articles. However, whenever articles presenting relatively slight but still visible color differences are involved, prior known color sensing methods which consider a limited number of wavelengths have failed to provide the resolution required. This is the case with color sensing of pieces made of a particular wood species and having to be classified in different categories for matching these pieces in wood furniture manufacturing. It has been found that red oak is a such species of wood, being characterized by a very high range of color hues, from light gray to dark red, and thus requiring a very high resolution color sensing and classification technique to classify in appropriate categories wood pieces giving substantially the same color perception to a person viewing at these pieces. Furthermore, high resolution has to be achieved while keeping the processing time required by the color sensing method as short as possible so as to implement such a method in a high volume automatic color article sensing apparatus, through minimizing the number of wavelength measurements and calculations.